Semi--Blind Robust Identification/Model (In)Validation with Applications to Macro--Economic Modelling. Pennsylvania State University
| Authors: | Sznaier Mario, Pennsylvania State University, United States
Ma Wenjing, Pennsylvania State University, United States
Yilmaz Muhittin, Pennsylvania State University, United States
Lagoa Constantino, Pennsylvania State University, United States |
|---|
| Topic: | 1.1 Modelling, Identification & Signal Processing |
|---|
| Session: | Model Validation Techniques |
|---|
| Keywords: | Robust Identification, Control Oriented Identification, Model (In)Validation, Macro-Economic Modelling. |
|---|
Abstract
This paper addresses the problems of worst--case identificationand model invalidation of systems subject to unknown initialconditions. While in principle these problems lead to non--convex BilinearMatrix Inequalities (BMIs), we show that tractableconvex relaxations are readily available. The potential of thesetechniques is illustrated by identifying and validating a subsystem of themacro--economy relating inflation and interest--rates. Theresulting model was able to correctly forecast the inflation rateover 149 quarters of real data,outperforming currently existing comparable linear models.
